Decision of Threshold Values for Extraction of Mangrove Area with TM Data of Landsat 5
It can be easily expected that the threshold value between waters and mangrove area exists in the right valley of the peak from water in both figures. In this case, searching the value may be easy between mangrove area and land area because the notch is clear for band 5. This matter changes with occupied ratio of three categories in the whole scene and condition of distribution of water in land area such as swamp, ponds and channels.
The way suggested here is to search fit threshold value by increasing or decreasing tentative value in the range of this valley or notch. A flow chart of this procedure is shown in fig.-2. The main procedure is to display five binary imagery by tentative values with reference imagery that is composed by RGB in a CRT. The fit threshold value is decided by synthetic judgment for fringe sharpness of categorized area, distribution of scattered dots in each side and change of total number of dots with increasing or decreasing K. inseparability between mangrove area and water or land area relates with the existence of cut over area, stand with reclaimed floor by crabs and squillae in mangrove area, swamp, ponds and channels in the land area. For removal of scattered dots in each side, logistic operation, filtering and topographic observation are useful. In fig.-2, procedure of operation between masks is shown for preparation of mask for mangrove area.
Fig . 2 Flow chart of procedure to decide fit threshold value between mangrove area and waters or land and making a mask for three categories
Results and Discussion
An example of display composed six imagery on a CRT for searching fit threshold value is shown in fig.-3 and fig.-4. If the fit threshold value is decided by drawing profiles of CCT count across boundary between mangrove area and water or land area, profiles must be drawn at many points across the boundary. It is not clear to set the point where and how many.
Fig.-3 An example of five binary imagery with reference composed by RGB in a CTR to search fit threshold value for making mask of waters with band 4
Fig.-4 An example of five binary imagery with reference composed by RGB in a CTR to search fit threshold value for making mask of land area with band 5.
